A method for setting multi-loop PID parameters in chemical process and its application

Abstract

The present invention discloses a method for tuning multi-loop PID parameters of a chemical process, which relates to the field of PID technology, and specifically includes the following steps: S1: initializing a particle swarm; S2: determining an optimization target and an evaluation function; S3: calculating the fitness of each particle according to the optimization target; S4: recording loop extreme values, individual extreme values, and global extreme values; S5: updating particle positions and speeds; S6: dynamically adjusting inertia weights and learning constants; S7: iterating steps S3 to S6 until the maximum number of iterations or termination conditions are reached. The present invention applies an improved multi-objective particle swarm algorithm to the multi-loop PID parameter tuning task of a chemical plant, and obtains that the comprehensive performance of the loop is improved by 3.68%. Compared with the original MOPSO algorithm, the convergence speed is improved by about 10%, and the use effect is effectively improved.

Description

Technical Field

[0001] The invention relates to the field of PID technology, and in particular to a method for setting multi-loop PID parameters of a chemical process and application thereof. Background Art

[0002] PID control is the most widely used control strategy in modern chemical process control. It is composed of the output of three parts: proportional, integral and differential. The coefficients of these three parts are called PID parameters. The setting of PID parameters directly affects the operation effect and quality of the control system.

[0003] However, there are some common shortcomings in the prior art of PID parameter tuning: due to the high-order, high-time-delay, and high-nonlinear characteristics of large chemical plants, the traditional empirical formula method is not effective in PID parameter tuning, especially in complex multi-loop systems. Since the dynamic characteristics and coupling degree of the system may vary, the general empirical formula is difficult to apply, for example:

[0004] In contrast, a method for parameter tuning of an internal model PID controller for multi-loop coupling in a chemical process with publication number CN106094910B analyzes the correlation between each loop by obtaining a relative gain matrix through sequential testing of multiple loops, and on this basis constructs a decoupling compensation matrix for the multi-loop system to decouple the multi-loop system, and finally uses the full-pole approximation method and the McLaughlin formula to calculate the PID controller parameters of each loop in turn. However, the full-pole approximation method and the McLaughlin formula used in this method to calculate the PID parameters cannot guarantee that these parameters are optimal.

[0005] Compared with the PID parameter automatic tuning method based on particle swarm algorithm with the publication number of CN115562221A, it combines the idea of ​​PID separation and maps the adjustment of PID parameters to the particle space; this method first initializes the parameters of the particle swarm, calculates the objective function value, then updates the individual and global optimal particle positions according to the objective function value, and finally realizes the parameter update through the PID mapping formula; this method can not only save the time of manual tuning, but also is expected to obtain better control performance than the traditional method. However, although the patent proposes a method based on particle swarm algorithm, it is only for single-loop systems. When it is extended to multi-loop situations, due to the large parameter space of multi-loop particle swarms, the convergence speed of the particle swarm algorithm may be extremely slow, which makes the tuning process consume a lot of time and affects the continuity of production.

[0006] Therefore, there is an urgent need to improve the above problems. Summary of the invention

[0007] The purpose of the present invention is to provide a method for adjusting multi-loop PID parameters in a chemical process and its application, so as to solve the problem that it is difficult to achieve good performance in the multi-loop PID parameter adjustment task of a chemical process proposed in the above background technology. And by improving the multi-loop PID parameter adjustment method based on the heuristic algorithm, its convergence speed is improved.

[0008] In order to achieve the above-mentioned invention object, the present invention adopts the following technical scheme:

[0009] The present invention provides a method for adjusting multi-loop PID parameters of a chemical process, comprising the following steps:

[0010] S1: Initialize the particle swarm; Initialize the particle swarm, that is, randomly assign an initial position and speed to each particle; For a chemical process section containing N PID control loops, the PID parameter of each loop is K n =(k Pn , k In ,k Dn ), where n represents the nth loop, the parameter space of the PSO algorithm can be defined as R （N×3） , randomly initialize I particles in the parameter space, and the position X of each particle i Represents a solution to the multi-loop PID tuning problem. Each particle is also assigned an initial velocity V i ;

[0011] S2: Determine the optimization objective and evaluation function; the objective function can be expressed as:

[0012] ;

[0013] ;

[0014] Where N represents the total number of control loops in the section, n represents the nth control loop, and a n represents the maximum overshoot of the nth control loop, r n represents the residual of the nth control loop, t n represents the transition time of the nth control loop, e n represents the energy consumption of the nth control loop, ε n Represents the weight of each loop, which is determined by the maximum overshoot of each loop;

[0015] S3: Calculate the fitness of each particle according to the optimization goal; perform weighted summation of the indicators of each loop according to the objective function to achieve the goal of optimizing all loops in the section at the same time; in order to follow the convention that the greater the fitness, the better the performance, take the inverse of the accumulated result to obtain the final objective function f(x);

[0016] S4: Record loop extreme values, individual extreme values, and global extreme values; the best position p experienced by each particle * i , called the individual extreme value, the best position g passed by the entire particle group * It is called the global extreme value. Based on the relatively independent characteristics of each control loop, the optimal position c of each control loop is added to the particle swarm algorithm. * , i.e., the loop extreme value;

[0017] S5: Update particle position and velocity; during the iteration of the algorithm, the particle updates its velocity V according to the following formula i and its position X i :

[0018] ;

[0019] Among them, V i ' represents the particle velocity calculated based on individual extreme values ​​and global extreme values, X i ' represents the new position of the particle after iteration; ω is called the inertia weight, which is used to control the weight of maintaining the original velocity of the particle, c 1 、c 2 and c 3 They are called local learning constant, global learning constant and loop learning constant respectively, which affect the speed at which particles approach the local optimal, global optimal and loop optimal positions respectively;

[0020] S6: Dynamically adjust the inertia weight and learning constant; inertia weight ω and each learning constant c 1 、c 2 and c 3 , so that it can be dynamically adjusted during the iteration process; the weight adjustment formula is as follows:

[0021] ;

[0022] Among them, ω max ,ω min 、c 2max and c 2min denote the inertia weight ω and the global learning constant c respectively 2 The maximum and minimum values ​​of , L is the total number of iterations, and z is the current number of iterations;

[0023] S7: Iterate step S3 to step S6 until the maximum number of iterations or the termination condition is reached.

[0024] Preferably, in step S4, when the multi-loop PID of the chemical process is adjusted, each loop has its own optimal position, and the optimal position of each single loop contacted by the particle group is spliced ​​together, that is, the optimal position of the loop.

[0025] Preferably, in step S6, the addition of the dynamic weight parameter allows a larger inertia weight ω to be maintained at the beginning of the iteration, so as to ensure that the particle swarm algorithm has good population diversity.

[0026] Preferably, in step S2, the evaluation index is obtained by calculating the response curve; the changes of the process variable PV, the control variable OP and the given value SP are plotted against time, and the resulting curve is the response curve; and the PID control loop performance is evaluated by using indicators such as maximum overshoot, residual and transition time.

[0027] Preferably, in step S2, the response curve is experimentally obtained on a real or simulated control system.

[0028] Preferably, in step S2, the maximum overshoot refers to the difference between the maximum deviation of the system response and the final value during the transition process.

[0029] Preferably, in step S2, the residual refers to the deviation between the new steady-state value reached by the controlled variable PV and the given value SP when the transition process ends.

[0030] Preferably, in step S2, the transition time refers to the time taken for the transition process from the moment when the interference occurs until the system re-establishes a new balance.

[0031] The present invention also provides an application of the above-mentioned chemical process multi-loop PID parameter tuning method, comprising the following steps:

[0032] A1: When the particle swarm algorithm records extreme values, a loop optimal extreme value is introduced; the optimal positions of each single loop touched by the particle swarm are spliced ​​together, that is, the optimal position of the loop, to form a set of optimal loop positions to guide the movement of particles, so as to accelerate the convergence speed of the algorithm and improve the stability of the algorithm;

[0033] A2: When the particle swarm algorithm updates the position and speed, by dynamically adjusting the inertia weight, local learning constant, global learning constant and loop learning constant, the particle swarm algorithm can better balance the global search and local convergence capabilities at different iteration stages, thereby improving the performance and efficiency of the algorithm.

[0034] Compared with the prior art, one or more of the above technical solutions have the following beneficial effects:

[0035] The present invention applies the improved multi-objective particle swarm algorithm to the multi-loop PID parameter tuning task of a chemical plant, and finds that the comprehensive performance of the loop is improved by 3.68%. Compared with the original MOPSO algorithm, the convergence speed is increased by about 10%, and the use effect is effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0037] Figure 1 is a schematic diagram of the motion state of particle i of the present invention;

[0038] Figure 2 It is a schematic diagram of the maximum overshoot in the PID loop performance evaluation index of the present invention;

[0039] Figure 3 It is a schematic diagram of the transition time in the PID loop performance evaluation index of the present invention;

[0040] Figure 4 It is a comparison diagram of the response curves before and after the loop PID parameters are adjusted in the present invention. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0042] This paper proposes an improved multi-objective particle swarm optimization algorithm (PID-MOPSO) to solve the problem of slow convergence of multi-loop PID parameter tuning tasks in chemical processes. By introducing the optimal position set of the control loop and optimizing the weight parameters, the convergence speed and performance of the algorithm can be improved.

[0043] This algorithm solves the problem of multi-loop PID parameter tuning in a chemical process. The input includes parameters related to the algorithm, the number of particles (I) of the particle swarm algorithm, inertia weight (ω), local learning constant (c1), global learning constant (c2), loop learning constant (c3), as well as the number of iterations and termination conditions of the algorithm. The fitness of the particles needs to be calculated from the response curve obtained experimentally on the real or simulated control loop, so the input of the algorithm may also include a simulation model of the control loop. The output of the algorithm is the solution set of the PID parameters of each loop after optimization and adjustment. The details are as follows:

[0044] A method for adjusting multi-loop PID parameters of the chemical process described above comprises the following steps:

[0045] S1: Initialize the particle swarm. Initialize the particle swarm, that is, randomly assign an initial position and velocity to each particle. For a chemical process section containing N PID control loops, the PID parameters of each loop are K n =(k Pn , k In ,k Dn ), where n represents the nth loop, the parameter space of the PSO algorithm can be defined as R （N×3） , randomly initialize I particles in the parameter space, and the position X of each particle i Represents a solution to the multi-loop PID tuning problem. Each particle is also assigned an initial velocity V i ;

[0046] S2: Determine the optimization objective and evaluation function. The objective function can be expressed as:

[0047] ;

[0048] ;

[0049] Where N represents the total number of control loops in the section, n represents the nth control loop, and a n represents the maximum overshoot of the nth control loop, r n represents the residual of the nth control loop, t n represents the transition time of the nth control loop, e n represents the energy consumption of the nth control loop, ε n Represents the weight of each loop, which is determined by the maximum overshoot of each loop;

[0050] S3: Calculate the fitness of each particle according to the optimization goal. Perform weighted summation of the indicators of each loop according to the objective function to achieve the goal of optimizing all loops in the section at the same time. In order to follow the convention that the greater the fitness, the better the performance, the reciprocal of the accumulated result is taken to obtain the final objective function f(x);

[0051] S4: Record loop extrema, individual extrema and global extrema. The best position p experienced by each particle * i , called the individual extreme value, the best position g passed by the entire particle group * Based on the relatively independent characteristics of each control loop, the optimal position c of each control loop is added to the particle swarm algorithm. * , i.e., the loop extreme value;

[0052] S5: Update particle position and velocity. During the iteration of the algorithm, the particle updates its velocity V according to the following formula: i and its position X i :

[0053] ;

[0054] Among them, V i ' represents the particle velocity calculated based on individual extreme values ​​and global extreme values, X i ' represents the new position of the particle after iteration. ω is called the inertia weight, which is used to control the weight of maintaining the original velocity of the particle. 1 、c 2 and c 3 They are called local learning constant, global learning constant and loop learning constant respectively, which affect the speed at which particles approach the local optimal, global optimal and loop optimal positions respectively;

[0055] S6: Dynamically adjust the inertia weight and learning constant. Inertia weight ω and each learning constant c 1 、c 2 and c 3 , so that it can be dynamically adjusted during the iteration process. The weight adjustment formula is as follows:

[0056] ;

[0057] Among them, ω max ,ω min 、c 2max and c 2min denote the inertia weight ω and the global learning constant c respectively 2 The maximum and minimum values ​​of , L is the total number of iterations, and z is the current number of iterations;

[0058] S7: Iterate step S3 to step S6 until the maximum number of iterations or the termination condition is reached;

[0059] Furthermore, in step S4, when the multi-loop PID of the chemical process is tuned, each loop has its own optimal position, and the optimal position of each single loop that the particle swarm contacts is spliced ​​together, that is, the optimal position of the loop. In the early stage of iteration, the optimal position of the loop may be closer to the global optimal solution than the global optimal position of the particle swarm. The introduction of the optimal position of the loop speeds up the convergence speed of the algorithm and improves the stability of the algorithm;

[0060] Furthermore, in step S6, the addition of dynamic weight parameters keeps a large inertia weight ω at the beginning of the iteration to ensure that the particle swarm algorithm has good population diversity. As the iteration enters the middle stage, the inertia weight ω gradually decreases, and the local learning constant c 1 and the loop learning constant c 3 The value of increases, so that particles can quickly gather around several local optimal points. At the end of the iteration, the local learning constant c 1 and the loop learning constant c 3 Rapidly decreases, the global learning constant c 2 Increase, thereby helping particles to jump out of the local minimum point. This improved weight update method improves the convergence speed of the algorithm and enhances the global search capability.

[0061] Furthermore, in step S2, the evaluation index needs to be obtained by calculating the response curve. The PID control loop automatically controls the process variable PV in the loop, so that when the process variable deviates from the given value SP due to the influence of disturbance, it automatically generates a control action to overcome the disturbance, and after a period of time, the process variable returns to the vicinity of the given value. The changes of the process variable PV, the control variable OP and the given value SP are plotted against time, and the resulting curve is called the response curve. The response curve shows the transition process of the control system to resist disturbance and restore the steady state, and is the basis for measuring the quality of the control system. Indicators such as maximum overshoot, residual and transition time are usually used to evaluate the performance of the PID control loop.

[0062] Furthermore, in step S2, the response curve needs to be obtained experimentally on a real or simulated control system.

[0063] Furthermore, in step S2, the maximum overshoot refers to the difference between the maximum deviation of the system response and the final value during the transition process. During the decay oscillation process, the maximum overshoot is the peak of the first wave. Figure 2 In the equation, the maximum overshoot (B) is the difference between the maximum deviation (A) and the final value (C).

[0064] Furthermore, in step S2, the residual is the deviation between the new steady-state value reached by the controlled variable PV and the given value SP when the transition process ends, or the residual is the residual deviation when the transition process ends.

[0065] Furthermore, in step S2, the transition time is the time taken for the transition process from the moment the disturbance occurs until the system re-establishes a new equilibrium. Generally, a small range is set above and below the steady-state value (generally set as ±2% of the new steady-state value). When the controlled variable enters this range and does not go beyond it, it is considered that the controlled variable has reached the new steady-state value. Figure 3 As shown in Figure 1, the time ST from the controlled variable entering the stable range for the first time and no longer changing to the start of the response is the transition time.

[0066] An application of a multi-loop PID parameter tuning method for a chemical process includes the following steps:

[0067] A1: When the particle swarm algorithm records extreme values, a loop optimal extreme value is introduced. The optimal positions of each single loop touched by the particle swarm are spliced ​​together, that is, the optimal position of the loop, to form a set of optimal loop positions to guide the movement of particles, so as to accelerate the convergence speed of the algorithm and improve the stability of the algorithm;

[0068] A2: When the particle swarm algorithm updates the position and velocity, by dynamically adjusting the inertia weight, local learning constant, global learning constant and loop learning constant, the particle swarm algorithm can better balance the ability of global search and local convergence at different iteration stages, thereby improving the performance and efficiency of the algorithm.

[0069] Based on this, the present invention conducts the following experiments for the above method and its application:

[0070] The section used in the following experiment comes from a methanol distillation unit in a certain factory. The unit adopts a three-tower methanol distillation process, which mainly includes three parts: a pre-distillation tower, a pressure tower, and an atmospheric tower. Four control loops are selected in the section, and the control loop table is as follows:

[0071]

[0072] According to the multi-objective particle swarm algorithm (PID-MOPSO) for PID tuning optimization of chemical processes described above, this experiment selected four loops, FET0705, FET0711, LT0702 and LT0712, for PID parameter search optimization. In the experiment, the number of particles was set to 500, the initial inertia weight ω was set to 0.7, and the local learning constant c 1 , global learning constant c 2The initial value of the loop learning constant c3 is set to 0.1. The evaluation of loop performance is carried out on the simulation model of this section. The experiment was repeated 30 times. The PID parameter tuning results of each loop are shown in the following table:

[0073]

[0074] In order to verify the control effect of the PID parameters obtained by the PID-MOPSO algorithm, a loop response experiment simulating step disturbance was carried out on the digital twin model of the chemical process based on the attentive graph convolutional neural network ASTGCN. In this experiment, the interference signal was introduced at the 50th time step. The obtained loop response curve is compared with the response curve obtained by the factory default PID parameters (1, -, -) to evaluate the performance improvement of the tuned PID parameters. The comparison results are shown in the table below, which shows the comparison of performance indicators before and after PID tuning on different loops, such as overshoot, steady-state error and response time. In addition, a comparison chart of the response curves of each loop before and after PID tuning is also drawn, as shown in the figure below. Figure 4 The response curve intuitively shows the impact of PID parameter tuning on loop response.

[0075]

[0076] pass Figure 4 It can be observed that the adjusted PID parameters are basically better than the default PID parameters in terms of residual, transition time and maximum overshoot. This shows that the PID parameter tuning algorithm based on the improved multi-objective particle swarm can improve the control performance of the loop. Compared with the default PID parameters, the adjusted PID parameters perform better in reducing residual, shortening transition time and reducing maximum overshoot. It can also be intuitively observed from the figure that in terms of steady-state error, the adjusted PID parameters can stabilize the system output near the given value faster. In addition, the adjusted PID parameters also show better performance in suppressing oscillation and stabilizing the control system output.

[0077] In summary, in the multi-loop PID parameter tuning task of a chemical plant, after applying the improved multi-objective particle swarm algorithm (PID-MOPSO) proposed in the present invention, the comprehensive performance of the loop is improved by 3.68%. Compared with the original MOPSO algorithm, the convergence speed is increased by about 10%, and the use effect is effectively improved.

[0078] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for setting multi-loop PID parameters in a chemical process, characterized in that: The following steps are involved: S1: Initialize the particle swarm; Initialize the particle swarm, that is, randomly assign an initial position and speed to each particle; For a chemical process section containing N PID control loops, the PID parameter of each loop is K n =(k Pn , k In , k Dn ), where n represents the nth loop, the parameter space of the PSO algorithm can be defined as R （N×3） , randomly initialize I particles in the parameter space, and the position X of each particle i Represents a solution to the multi-loop PID tuning problem. Each particle is also assigned an initial velocity V i ; S2: Determine the optimization objective and evaluation function; the objective function can be expressed as: ； ； Where N represents the total number of control loops in the section, n represents the nth control loop, and a n represents the maximum overshoot of the nth control loop, r n represents the residual of the nth control loop, t n represents the transition time of the nth control loop, e n represents the energy consumption of the nth control loop, ε n Represents the weight of each loop, which is determined by the maximum overshoot of each loop; S3: Calculate the fitness of each particle according to the optimization goal; perform weighted summation of the indicators of each loop according to the objective function to achieve the goal of optimizing all loops in the section at the same time; in order to follow the convention that the greater the fitness, the better the performance, take the inverse of the accumulated result to obtain the final objective function f(x); S4: Record loop extreme values, individual extreme values, and global extreme values; the best position p experienced by each particle * i , called the individual extreme value, the best position g passed by the entire particle group * It is called the global extreme value. Based on the relatively independent characteristics of each control loop, the optimal position c of each control loop is added to the particle swarm algorithm. * , i.e., the loop extreme value; S5: Update particle position and velocity; during the iteration of the algorithm, the particle updates its velocity V according to the following formula i and its position X i : ； Among them, V i ' represents the particle velocity calculated based on individual extreme values ​​and global extreme values, X i ' represents the new position of the particle after iteration; ω is called the inertia weight, which is used to control the weight of maintaining the original speed of the particle; c1, c2 and c3 are called the local learning constant, the global learning constant and the loop learning constant, respectively, which affect the speed at which the particle approaches the local optimal, the global optimal and the loop optimal position respectively; S6: Dynamically adjust the inertia weight and learning constant; the inertia weight ω and the learning constants c1, c2 and c3 are dynamically adjusted during the iteration process; the weight adjustment formula is as follows: ； Among them, ω max ,ω min 、c 2max and c 2min They represent the maximum and minimum values ​​of the inertia weight ω and the global learning constant c2, respectively, L is the total number of iterations, and z is the current number of iterations; S7: Iterate step S3 to step S6 until the maximum number of iterations or the termination condition is reached.

2. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S4, when the multi-loop PID of the chemical process is tuned, each loop has its own optimal position, and the optimal position of each single loop contacted by the particle group is spliced ​​together to obtain the optimal position of the loop.

3. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S6, the addition of the dynamic weight parameter allows a larger inertia weight ω to be maintained at the beginning of the iteration, so as to ensure that the particle swarm algorithm has good population diversity.

4. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S2, the evaluation index is obtained by calculating the response curve; the changes of the process variable PV, the control variable OP and the given value SP are plotted against time, and the resulting curve is the response curve; and the PID control loop performance is evaluated by using the maximum overshoot, residual and transition time indicators.

5. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S2, a response curve is experimentally obtained on a real or simulated control system.

6. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S2, the maximum overshoot refers to the difference between the maximum deviation of the system response and the final value during the transition process.

7. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S2, the residual refers to the deviation between the new steady-state value reached by the controlled variable PV and the set value SP when the transient process ends.

8. The method for setting multi-loop PID parameters of a chemical process according to claim 1, characterized in that: In step S2, the transition time refers to the time taken for the transition process from the moment when the disturbance occurs to the moment when the system re-establishes a new equilibrium.

9. An application of the multi-loop PID parameter tuning method for a chemical process according to any one of claims 1 to 8, characterized in that: The following steps are involved: A1: When the particle swarm algorithm records extreme values, a loop optimal extreme value is introduced; the optimal positions of each single loop touched by the particle swarm are spliced ​​together, that is, the optimal position of the loop, to form a set of optimal loop positions to guide the movement of particles, so as to accelerate the convergence speed of the algorithm and improve the stability of the algorithm; A2: When the particle swarm algorithm updates the position and speed, by dynamically adjusting the inertia weight, local learning constant, global learning constant and loop learning constant, the particle swarm algorithm can better balance the global search and local convergence capabilities at different iteration stages, thereby improving the performance and efficiency of the algorithm.

Images